Many of your clients have been quietly turning to ChatGPT before calling you.
They type in a question about their tax situation, get a detailed-sounding response, and either act on it or arrive at your desk with assumptions already formed. But that will no longer be the case now.
On October 29, 2025, OpenAI updated its usage policy, which directly affects how ChatGPT handles financial and tax-related questions. Here is a clear breakdown of what the update actually says, what it does not say, and what it means for professionals in tax, accounting, and bookkeeping.
What OpenAI Actually Changed
OpenAI’s updated Usage Policy reclassifies ChatGPT from an advisory tool to an educational one. The policy now explicitly prohibits the platform from providing personalized financial, legal, or medical advice, or being used for consultations that would normally require a licensed professional.
In practical terms, ChatGPT will no longer offer personalized tax guidance, financial planning recommendations, or investment decisions tied to a user’s specific situation. It also cannot support high-stakes decisions in areas such as employment, housing, or migration without clear human oversight.
The change is driven by regulatory pressure and liability concerns. At the same time, documented cases of real harm have added urgency.
- A patient was hospitalised after replacing table salt with sodium bromide based on ChatGPT advice.
- A man delayed seeking medical help after the tool suggested cancer was unlikely and was later diagnosed at stage four.
Tax and financial decisions carry the same potential for serious consequences when AI-generated responses are treated as professional guidance.
What ChatGPT Can Still Do
It is important not to overstate the scope of this change. ChatGPT has not gone silent on financial and tax topics. The platform can still explain general concepts clearly. It can describe how tax brackets work, define what a capital gain is, explain the difference between a sole trader and a limited company, or summarize what a particular regulation covers in broad terms.
What it can no longer do is apply that information to a specific person’s circumstances and present the output as guidance. The line being drawn is between general information and personalized advice. A user asking “how does depreciation work” will still get a useful answer. A user asking “Should I claim this expense given my income and business structure?” should, under the new policy, be redirected to a qualified professional rather than given a direct answer.
This distinction matters for tax professionals to understand because it means ChatGPT remains a legitimate general information tool. It has not become useless in this space. It has simply been formally moved out of the advisory lane.
What These ChatGPT Restrictions Mean for Your Clients
The most direct impact on your clients is simple: a tool many of them have been using as a free substitute for professional advice has now officially stopped playing that role. Clients who were getting personalized-feeling tax and financial responses from ChatGPT will increasingly be redirected to seek qualified help instead.
That said, the shift is gradual rather than immediate. Enforcement has limits, and clients who push with specific scenarios may still get responses that feel advisory. But the policy puts OpenAI on record, and that changes the accountability conversation. If a client acts on AI-generated tax guidance from this point forward, they are doing so against the explicit position of the platform itself.
The confidentiality angle is also worth raising with clients directly. Conversations with ChatGPT carry no equivalent of adviser-client privilege. They can potentially be subpoenaed. For clients who have been sharing financial details with an AI chatbot, that is a material risk they may not have considered.
What These ChatGPT Restrictions Mean for Your Practice
On the client intake side, some people who were previously self-serving with AI on tax and financial questions may now be more inclined to seek professional input. The policy does not guarantee a shift in behaviour, but it does remove a layer of confidence that many users placed in AI-generated responses.
It also gives practices a clearer way to communicate their value. A licensed tax adviser is regulated, insured, and professionally accountable for the guidance they give. An AI chatbot, regardless of how credible its output sounds, is none of those things. This has always been true. The OpenAI update simply makes it easier to say out loud.
The third point is worth being clear on internally. This policy restricts what ChatGPT tells clients. It does not restrict how practitioners use AI in their own workflow. Using AI to draft documents, organise information, or research general positions is entirely unaffected. The line being drawn is between a professional using AI as a tool and a client using AI as a replacement for the professional. Those are two different things, and this update makes that distinction official.
The Bigger Picture
OpenAI’s policy change reflects a broader trend in how regulators are approaching AI in high-stakes fields. Governments across multiple jurisdictions are moving toward requiring greater transparency, accountability, and human oversight in AI systems that touch decisions with serious real-world consequences. Financial and tax advice sit clearly within that category.
For the AI industry, this represents a calibration period. The initial wave of adoption was relatively permissive. The current period involves a more careful reckoning with where AI can operate without adequate safeguards and where it cannot. Professional advice in finance, tax, and law is an area where regulators and companies alike are concluding that human oversight is not optional.
This is also unlikely to be the last update of this kind. As regulatory expectations tighten, AI platforms operating near professional services will face continued pressure to clarify and enforce their limitations more rigorously.
Conclusion
OpenAI’s October 2025 update does not change what licensed professionals do. It simply confirms, officially, that AI cannot do it.
For tax professionals, the distinction between what AI can do and what a qualified professional does has now been acknowledged by one of the largest AI companies in the world. That distinction has always existed. It is now simply a matter of record.
